On this page
Troubleshoot a pipeline¶
Start with the symptom, then use the full error title and context. Several LHP error codes are used by more than one validation path, so the code alone may not identify the cause. The error catalogue has further explanations.
A pipeline or flowgroup is missing¶
Check pipeline and flowgroup names and the project’s include,
blueprint_include and instance_include patterns. Check any CLI pipeline
filter. Run lhp validate --env dev with the same selection as generation;
in a bundle project, also pass -pc config/pipeline_config.yaml.
See lhp.yaml: project configuration and Flowgroup YAML.
A substitution token does not resolve¶
Check the selected --env, the filename substitutions/<env>.yaml and the
mapping with that environment’s name inside the file. Secret aliases live in
the top-level secrets mapping. Blueprint/local %{...} variables and
environment ${...} tokens have different scopes. See
Substitutions and Blueprints.
Bundle generation asks for pipeline configuration¶
When databricks.yml exists, supply -pc config/pipeline_config.yaml to
both validation and generation. To generate without bundle resources, pass
--no-bundle. Check that each selected pipeline resolves a catalog and schema.
See Configure and deploy bundles and pipeline_config.yaml.
Test actions did not appear¶
Use --include-tests when generating test actions. If results are not being
published, check test_reporting in lhp.yaml and the provider’s function
and configuration file. Provider fields belong to the project; test_id
belongs to an individual test action. See Test reporting configuration
and Report test results to an external system.
Monitoring tables are empty or the job fails¶
Check that both event_log and monitoring are enabled, the source pipelines
have run and emitted logs, and the monitoring job was deployed and run. Check
the run-as identity’s access to source logs, output tables and checkpoint storage.
Inspect the union-notebook task before the summary-pipeline task. Use
Monitoring configuration for names and inheritance and
Monitoring job configuration for job settings.
A dependency is missing¶
Inspect the graph with See how your pipelines connect. Add depends_on
when a dependency cannot be inferred from source code, and check whether
--trust-depends-on changes the analysis. See
Dependency analysis reference for matching and scope rules.